The Startup Ideas Podcast - "Y Combinator Is Overrated" - Inside the brain of this 20-Year-Old Who Built Multiple Viral Products

Starting point is 00:00:00 This is one of our more crazy startup ideas podcast episodes. It's with Avi Schiffman, the founder of Friend.com. He was all over the news for raising a few million dollars and spending most of it on buying a domain. He's got some spicy takes. You know, if you view your work as art, you have no competition and you have no fear or failure. Like move, move, yeah, like move fast and break things, I think is the thing I hate the most. I mean, I think maybe it works for shitty software, but like, I don't. think, let's say in my situation, I'm trying to build hardware.

Starting point is 00:00:32 You know, I only really get like one real first impression. Any business really does. Yeah, you know, no one great ever thought they couldn't be great. You know, you really need to have instilled in yourself a deep-rooted self, you know, just, you know, you need to just believe you are the listener guide in a sense. And I think will take you further than anything you will ever hear from someone else. He also gives away a free startup idea in the AI space. It's highly technical. So many of you might not understand it,

Starting point is 00:01:05 but for those of you who do, he gives a playbook on how to create this technical AI app that he thinks could make millions. You might not agree with everything Avi says, but it makes for a very interesting conversation and it probably will get your creative juices flowing. We got Avi Schiffman on the pod, founder of friend.com,

Starting point is 00:01:48 22 years old, Harvard Dropout. He makes a lot of noise on Twitter and places like that. So thanks for coming on, man. For sure. Thanks, Craig. Yeah, it's cool to be on a podcast

Starting point is 00:01:59 to just talk about ideas, I guess. Big idea guy here, of course. I think I definitely have large tunnel vision. I think what I usually think about maybe more is like why people are usually failing at their startups or why a lot of people's ideas

Starting point is 00:02:15 don't work or like... Yeah. Let's talk about that. Yeah, I think that a lot of, I mean, there's like a very famous maxim, right, where like most people's startups die from suicide rather than competition. And I, you know, I really believe. I mean, let's say, let's say in my situation, right, like I've been building this AI pendant, okay, this AI wearable pendant for a while now. I started this company back when it's called Tab, right, in May of May of last year in 2023. And since then there's been a gazillion, you know, large companies and random kids and idiots and all, ever, ever. everyone all the like all trying to build their own version of this. And many more than half of the ones that I remember being started have all failed or, you know, not, you know, maybe I'm not going to be too harsh on them. Although, you know, they're trying to copy me still. But, you know, they moved on to like other ideas. But, you know, if I was so worried about all those competitors, which I never once was, you know, I think that would have ruined my startup a lot more. Maybe I would have eventually, you know, killed my own startup. I think people are far too worried about that. And I think that. And I think that the true reason, but the true way that you get over that is if you truly have confidence in yourself where you know that you are entirely acting off your own intuition, well, I guess, you know, to entirely act off your own intuition, again, you really do need to be confident in your

Starting point is 00:03:34 abilities as an idea guy, I guess. But, you know, if you view your work as art, you have no competition and you have no fear or failure. And I think it just removes this. anxiety that you have over it needing to be perfect or anything like that. Well, I mean, at the same time, like, you know, it's, it's not perfect. It may be in other people's eyes. I think it's perfect to your eyes, which might be a bit harder, but, you know, I think that's the way to go. I also think that Y Combinator is evil and has ruined startups. I think it's turned it into just absolute slop. And, like, move, yeah, like, move fast and break things, I think is the thing I hate the most. I mean, I think maybe.

Starting point is 00:04:18 it works for a shitty software, but like, I think, let's say in my situation, I'm trying to build hardware, you know, I only really get like one real first impression. Any business really does. And like, you know, do you think like the guy painting the Mona Lisa was like, oh, you know, okay, it's, you know, it's good enough. It was never, no artist has ever said their work is good enough. And I, I think, you know, there is a line you, you, you know, you shouldn't, you know, go too far away from. But, um, I think people should just ship when they, when they, when they shit and uh you know maybe if you do also want to kill your startup then kill it but uh i think like just truly freeing your mind away from competition like that um and truly just building what you

Starting point is 00:04:59 truly want to build and not being impacted by like i don't know this like wicy pocket advice book i see all these founders like pin to their uh yeah yeah i don't know i mean like i guess it's helpful for some people and i think like you know there's definitely a genre of startups these days where a lot people are building SaaS and it's very templated and I guess if you need to be, you know, have your handheld and go for it. But there was one more point I wanted to make. I think there's a huge difference between entrepreneurship and innovation. And I think this is something that a lot of people maybe get kind of in not unknowingly stuck in where like, I think a lot of what I'm saying maybe applies a lot more to people trying to build something

Starting point is 00:05:40 new and innovate and maybe in that sense it's a lot easier to view your work as art because you are truly creating something new. I think for a lot of people that are maybe not inherently motivated by fulfillment in that sense and are really just trying to make a quick buck, which, you know, if you want to do that, that's fine. I think it's probably fine to follow advice like that. What do you think of founders who aren't trying to reinvent the wheel, who, you know, they're trying to create a SaaS business that is five, ten million bucks a year. I just think, like, we're on the cusp of AGI. You're really going to spend your time doing that?

Starting point is 00:06:17 Like, there's only so many things you'll ever work on in your entire life. And I think what I hate the most about most people I meet around the world is everyone always is trying to prepare for something. You know, some people are trying to work at some startup for 10 years or whatever so they can gain startup experience or, you know, I mean, it's not even just in the tech industry and in our little scene. It's like just people out there that are a lot of people are maybe stuck in college or a lot of people just don't know what they want to do. And I think one that boils down to that's, you know, skill issue, I would say again, like it's your fucking problem if you haven't figured out what you want to do by then. And quick break in the pod to tell you a little bit about startup empire. So startup empire is my private membership where it's a bunch of people like me, like you, who want to build out. their startup ideas. Now they're looking for content to help accelerate that. They're looking for

Starting point is 00:07:11 potential co-founders. They're looking for tutorials from people like me to come in and tell them, how do you do email marketing? How do you build an audience? How do you go viral on Twitter? All these different things. That's exactly what startup empire is. And it's for people who want to start a startup but are looking for ideas or it's for people who have a startup, but just they're not seeing the traction that they need. So you can check out the link to startup empire.co in the description. Well, I mean, I think it's easier, you know, to create something new than it is to make something better.

Starting point is 00:07:48 And I think maybe it's scarier, let's say, to do something unknown for sure. But like, okay, like, let's say, let's say what I'm really trying to do, which is to build like an embodied AI friend. That is something new, rather than necessarily trying to build a better chatbot. I know I did launch friend.com, but kind of ignore that for a second. Right. Like, if I was truly trying to compete with building a chatbot, chatbot, like it's kind of existed since the 60s with Eliza, you know,

Starting point is 00:08:17 you're trying to build something better, right? You're trying to build a better user interface that looks nicer. And, you know, the chats are a little nicer maybe, which is kind of like what I went through, actually, you know, with building, let's say, friend.com, right? I'm trying to build something that is going to be compared against all these other web-based chatbots like replica and character AI and you know even let's say chat jubcuit writer or just all these all these chatbots but you know if i'm trying to build

Starting point is 00:08:42 um something entirely new that's like a embodied uh i i friend you just it's it's entirely new and there's nothing to really compare it with except i don't know maybe like a dog or like a tamagotchi or something like that um you know i don't really have to make it so so fancy i don't necessarily have to have the best user interface or any of these things. It's a bit easier to just have this novel thing that is, I don't know, like just, yeah, there's nothing to compare it to. And it stands on its own and it creates new categories. It's easier to create a new category than to try and, you know, win in an existing category. Because if you create your own category, you immediately crown yourself king. And I don't know why anyone builds in categories that already exists unless you're

Starting point is 00:09:32 like a cuck. You know, you're truly cucking yourself if you're, if you're building in a existing category. There's already a king. What would you say to people who would say that everything is an existing category? Like what you're building is in an existing category. I mean, I don't know. You can make different levels of scopes. Sure. I mean, yeah, I think there's an argument that yeah, all startups or startups or some bullshit like that. But like, I don't know what to tell you. I think there's still as clear differential like categories for sure. Well, maybe, you know, when you truly, it needs to truly be like a new category. Look, you can't just build a slightly better note-taking app on, you know, I don't know.

Starting point is 00:10:09 It's still a category of productivity even. And let's say the consumer social category. What is TikTok a new, in a new category or is it, you know, like, how would you see TikTok versus Instagram? You know, that's a good point, right? Like, there's definitely situations where you can dethrone a king of an existing category. I think TikTok is the only consumer social product that came out after, you know, the OG ones in the, in the earlier 2010s that actually succeeded in, I think, truly becoming king in a way. Like, I, that TikTok is great. But I think that's very hard.

Starting point is 00:10:50 I guess, like, yeah, I mean, like, you can go for it. You could build a new car. Like, let's say Tesla, right? Like, they entered, they are, okay, in some ways, slightly different category, but overall, it is still the same category of, like, cars. you know, they dethroned Ford and a lot of these other existing automobiles. But again, I feel like they still did something slightly different, right? Like, they still are entering a category of, let's say, automobiles, but they still are doing it in a large enough of a different way.

Starting point is 00:11:16 And I think there is a large enough of a difference between like ice cars and, you know, electric vehicles. And there is, you know, a note taking up than like a note taking app powered by AI or something. Like, that's not really enough of, a change, a categorical change, right? Yeah, I think what you're saying is kind of like try to, try to build and be a category of one, so to speak. Yeah, it's easier.

Starting point is 00:11:42 Is that, yeah. I don't know. I truly think it's easier because, again, you have back to like what I was mentioning earlier where people are so afraid of competition. I think, you know, when you create your own category, you have less competition in a sense because there's less people building in that space. And I would really encourage people to also zoom out a little bit with these categories, right? Like, let's say, you know, in my opinion, productivity as a category is over and boring and done.

Starting point is 00:12:10 And, you know, in Combrance, like Apple, Google, or, you know, even beyond technology startups in a sense, too, it's done. It's over. Like, build something an entirely new category. Like, to me, the new category that I'm really trying to truly spearhead and grow is just, like, digital companionship or just kind of companionship as a category overall. I don't think that, I think when most people have thought about technology, that has always been very utilitarian and, you know, not, not focused on emotions and love and intimacy and companionship and like these things.

Starting point is 00:12:44 Oh, and one point I always like to mention is, you know, I think that the only real moat is consumer mindshare or just, you know, mindshare as a concept of, you know, When people think of a certain category or topic, their mind immediately goes to just one project or one company, right? Like, for example, if you think of chat chabit versus Claude, I use Claude. You know, I don't use Chatsbytee anymore. Maybe you're the same. I'm sure you've seen on Twitter and are seeing almost everyone loves Claude, right, more than ChachypT. But if you look at like, let's say, Google search trends, JCP is literally at like 100 and Claude is literally at zero, right?

Starting point is 00:13:19 Like, Claude is, an anthropic is insignificant. Nobody knows about them at all. they win on Mindshare more than they win on necessarily a better technical product or a better team or financials or anything like that. Mindshare is the only thing that matters, but it can be a very double-edged sword, right? Like when you think about, let's say Apple TV versus Netflix, when you think about Apple, you think of them as a computing or productivity company, right? You do not think of them as an entertainment company. But when you think of Netflix, you think of them as a, you know, entertainment streaming company, right? And I think that's why Netflix will still always beat something like Apple TV or, you know, let's say in my situation, I know that meta or Facebook, whatever you want to call them has been venturing a lot into like AI companions and digital relationships. You know, they've got little AI friends on Messenger or they've got a character AI clone actually on Instagram on the web version of it. But when you think of meta, you know, I think you truly think of like human to human relationships. And I think it's a very unique situation you're in as a startup to create like a new POV.

Starting point is 00:14:21 in my case, which is, I think, like, human-to-AI relationships. Like, that is what, you know, my mind-share concept is. And I think it's very, very hard for existing encumbrance to, like, as I said, it's a double-edged sword. You know, they might have a lot of mind-share in that category then, right. But then to, like, really change that and build from a new POV, I think it's really, you know, it takes really strong marketing campaigns over a long period of time. You know, like, let's say, let's say, I don't know, chat to be to you.

Starting point is 00:14:51 is inherently known as utilitarian and productivity, but I think Sam maybe has a hard on for still trying to build, like, her, or maybe her is still kind of an assistant, but like, let's say a true AI girlfriend type of thing. I just think they would really struggle with that ultimately because, like, their brand is so solidified and being known as productivity. And I think this is the only real advantage startups have.

Starting point is 00:15:13 I always hear people say that your advantage of a startup is that, like, you know, you can be focused or, like, you can move fast, or like, you know, you're competing, I guess, maybe against just a small team at Google, even though they're a giant company. And, like, sure, I guess these are kind of only skill issue reasons. But, like, the only real thing that you have an advantage of is a fresh slate in, like, people's minds, right, of, like, what your company's values are and what your, what your category, I guess, in a sense, really is.

Starting point is 00:15:41 I think that is also only really applicable if you're trying to build something big and real. But, you know, again, still, I think the real. the real advantage you have, right? Could you create, you know, a new category without raising venture capital or is this something that you need to raise money to go and do? You probably could without raising venture capital. I don't know, maybe you're really rich already. You can bootstrap it or, you know, I don't know, maybe you've got some sketchy deal

Starting point is 00:16:11 go. I don't know, sure. There's definitely, it's definitely not an absolute. Yeah, it's definitely not an absolute that you need to raise VC capital. VC capital is definitely useful, you know, to spend on domains and stuff like that for sure. But, you know, I think if you care about innovation instead of entrepreneurship in a sense, and you really are trying to build something to last a long time, I think you should care a lot about, you know, your equity.

Starting point is 00:16:36 And, you know, be careful about that to not raise too much venture capital. But, you know, you don't need to raise $100 million. You can still raise like $5 million or $250,000. right? You don't need to go extreme with it. Cool. Yeah. I ask just because a lot of people listening here,

Starting point is 00:16:56 a lot of people, I presume you live in San Francisco or the Bay Area. Just like a lot of people listening here. That world is so alien to them. You know, like they, yeah, venture capital and stuff like that. So a lot of people listening are just building products. bootstrapping at themselves. And so, you know, just question.

Starting point is 00:17:23 Yeah, yeah. I mean, I think the topic of like raising a VC capital for people that are not, you know, and so networked as well as I guess you are, I don't know where you live. L.A. New York. I'm in Miami. Okay. Fancy fancy fancy. But yeah, you know, I do live in San Francisco and I am on, on, you know, tech Twitter quite a lot.

Starting point is 00:17:41 And, you know, I'm very networked, of course. But, like, ultimately, I think to raise venture capital, The more you talk about your product and what you're working on, I actually think the harder it really is. Because I don't think any seed or lower investor gives the fuck about what you're working on. I think it's almost entirely, you know, investing in you and your ability to yap. And therefore, I think it's honestly quite easy to raise VC in a sense. You truly just need to be good at yapping. I also think that the real way you raise is when it becomes about the deal and not about what you're actually working on in a sense.

Starting point is 00:18:17 I don't know how to explain that particularly, but just like everything, VC is a social game. And these VCs are inherently just kind of trying to get a better, you know, they're competing against their, their coworkers and their friends at other funds. And like, that's what they really care about. They care about who's involved in certain deals. You know, when your fundraise is about the deal and not about your product, I guess, or your company, that's, I think, when it becomes so easy. How to get to that point?

Starting point is 00:18:46 I don't know. You either know how to do it or you don't. It's not like anyone will really give you some specific advice like that. All I think that the only thing that people should really be optimizing for is personal confidence. I think it bleeds into everything from how good you are at yapping, how good you are at recruiting, how good you are at following your own intuition, which makes it easier to make, you know, major product decisions or just any decisions at all. And, you know, how also it removes your fear of competition, I would say strongly. I truly think that, like, if I succeed in stuff one day, like, I really want to write a lot more and instill more of this underlying sense of personal confidence is what most founders should be working towards rather than, like, why see advice? Because if you are confident enough in yourself, and again, you need to act, like, I think the thing is that you still need to actually have good ideas and you need to be smart.

Starting point is 00:19:37 But like, if that's where you're at, I have no advice to give you. That's your fucking problem. But assuming you meet, you know, the bare, you know, bar minimum requirements. you should be maximizing to just believe in you or self having good ideas, essentially. And if you have that, you don't need to follow advice from White Compinator or whatever I'm saying to you as a startup or anything like that because you just, you know, you believe in yourself. And I think, you know, not there's a difference between arrogance and confidence in a sense, although I think it's only arrogance in hindsight is something I believe in for sure.

Starting point is 00:20:12 I think a lot of people probably think I'm arrogant, but I think that's only, you know, until I succeed, I suppose, then it's viewed as confidence. I would not be bothered by those perceptions, though, if I was someone. But, yeah. How does someone gain confidence, you know? Is that something that, you know, you can learn or? I think your VC funding should be spent on trips to IBs or something. No, not seriously, you know, but I think a lot of people should be focusing a lot more

Starting point is 00:20:42 on their personal life outside of their work because it ultimately actually ends up improving their work life actually quite significantly. I feel like this is also a much more strong point I'm trying to hit home for I think probably more so younger founders. Again, I'm 22. When I started working on this project, I think I'd only recently turned 21. This is the first time I've even, you know, lived in my own real place outside of home. And like, you know, I'm basically still a, you know, and you're all so weird if you're in tech. Like I'm, you know, I think you just, I've, I've grown so much from my own personal confidence and my own, you know, sense of self.

Starting point is 00:21:24 And I think that's because I've spent so much time not necessarily working, although you are kind of working, maybe if you view it that way, to just, like, travel a little bit, like, go do things outside of your comfort zone a little bit. I mean, I think that, that phrasing is cringe. I don't really know what to tell people. Like, earlier this year, I, for example, I was in Berlin at one point. And I went to like Barakhin actually with one of my prototypes.

Starting point is 00:21:50 Crazy, you know, it's crazy experiences. But like I think these are the kind of things that have instilled maybe like just it's cool. It's confidence. It makes you grow as a person. And as a result, I've noticed that 100% bleed into how good I am. Maybe had fundraising or hiring or, you know, convincing people of ideas. And, you know, I think that's had more of an impact and any advice someone's given me or anything I've read online or, you know, just really anything else at all.

Starting point is 00:22:20 I don't even really know why not many people talk about this. Because any founder that people look up to, or it doesn't have to be founder, you know, artists, let's say the Beatles or something like that. They, that is like their number one skill in a sense is that they're very confident in themselves. And that's, you know, you think you cross a threshold of confidence where everything you do really does become an expression of who you are because it's so in tune with who you are that it truly is, you know, artwork. I think that is the ultimate intrinsically valuable thing to, to work on as well. And what will truly make you feel confidently proud about yourself, which I think is the optimal

Starting point is 00:22:56 emotion to truly feel. And yeah, you know, no one great ever thought they couldn't be great. You know, you really need to have instilled in yourself a deep-rooted self, you know, just, you know, You need to just believe you are the listener guide in a sense. And I think it will take you further than anything you'll ever hear from someone else. I saw this really good tweet by this guy, Aaron Alto. He tweeted, visualization is the first step. And I think that's true in a lot of ways. Like, you know, for startup ideas, you know, once, you know, step zero is visualization.

Starting point is 00:23:37 Or maybe step, yeah, maybe actually upon a problem. reflection. Maybe step, you know, step, step one is find your startup idea. Step two is visualization. Yeah, I think, you know, this concept of manifesting and stuff is actually real in a sense, right? You know, and I think maybe if you relate this to like your, again, like your personal self, a lot of people want to visualize like the kind of person they want to be, right? Like I know a lot of girls or something like that. Like tape photos of these kind of women that they want to end up being, right? and you just kind of embody that over time. You do need to kind of turn yourself into a, you know, schizophrenic.

Starting point is 00:24:18 But, you know, pulling you towards where you want to go. I will say, I think people that are, like, trying to come up with startup ideas, though, it's like, I don't really know how to approach saying this or, like, saying anything to say. But, like, I don't know. I feel like you just got to live and, like, ideas just kind of come to you. because, you know, okay, it's great to have, like, one startup idea, but, like, your idea will change. Like, let's say what I'm working on right now. It's friend, right?

Starting point is 00:24:48 It started out as tab, and it was originally made because I was fundraising for a nonprofit, and I wanted to keep track of the people I was meeting. So I felt that if I wore, like, an always-on microphone, you know, and use LLMs and transcribe all the data, you could, like, automatically build a person's schema, right? That is completely different than building an embodied AI friend. the underlying tech is actually still the same. You know, you will go down in IDMAs and you will make thousands and thousands of more ideas. You'll make a billion ideas all day every day.

Starting point is 00:25:16 I think you just need to, again, I would say work on your own self-confidence and do things in the real world. And when the real ideas come to you, you will feel confident enough to spring on them. And you will feel confident enough to actually go down the entire IDMAs and continue to believe in your own ideas and truly see it to fruition. And, you know, you need to have confidence to allow yourself to actually work on something for, you know, a year or two or how long it takes to like truly bring something to market. You think Sam Allman had no confidence in himself while he toiled away at Open AI for a billion years until they launched Chat. You know, I, yeah, I think, I think confidence is the real, basically, skill issue.

Starting point is 00:25:54 It is the clear biggest divide between anyone that does anything at all outside. You know, I'm not even talking about startups. I'm talking about again, like just, just anything worthwhile. It is the biggest blocker. I can't think of anyone that wasn't confident that went on to do something. thing that anyone cares about. That could be, that's actually a good tip. So, you know, for people listening, it's like, if you're, if you're feeling low confidence,

Starting point is 00:26:17 maybe ask yourself like, what would Sam Altman do in this case? If that's something to look up to you, right? Smoke some weed. I don't know, chill out a little bit. That's what I would recommend. There you go. I'm serious. I'm serious.

Starting point is 00:26:29 People are way to like, I mean, everyone's different. I honestly, this advice I'm giving is not for, maybe it's not really relevant. But I just implore people to do things in the real world and grow as a person outside of, you know, I'd say focus on real world relationships too. I think that is how you grow more than than anything in the world. You know, not just girlfriends, friends to, you know, family, things like that. You grow so much from these things. It is incredible. I think it's a great way to focus on basically your personal growth and maybe a slightly more structured way that can lead to basically what I'm saying.

Starting point is 00:27:04 Quick ad break. Let me tell you about a business. I invested in. It's called Boringmarketing.com. So a few years ago, I met this group of people that were some of the best SEO experts in the world. They were behind getting some of the biggest companies found on Google. And the secret sauce is they've got a set of technology and AI. I could help you outrank your competition. So for my own businesses, I wanted that. I didn't want to have to rely on Mark Zuckerberg. I didn't want to depend on ads to drive. customers to my businesses.

Starting point is 00:27:39 I wanted to rank high in Google. That's why I like SEO and that's why I use Boringmarketing.com and that's why I invested in it. They're so confident in their approach that they offer a 30-day sprint with 100% money-back guarantee. Who does that nowadays? So check it out. Highly recommend boringmarketing.com. There's some people who work 120 hours a week who don't leave their house and just work, work,

Starting point is 00:28:03 work work. And I'm not of that belief. I think that you come up with good ideas and clarity of thought when you work a bit more, when you're more balanced. Do you agree? Yeah, you just got to kind of like putter around honestly and like just kind of walk around in new environments especially and just kind of make sure that you're right. Like again, I think people think I work very hard. Maybe I give that impression off Twitter, although I don't think I really do. I'm mostly posting pictures and motorcycles, but, like, I feel like how I work hard is in trying to make sure that I'm going to work hard for the right thing, right? Like, I think there was a famous, I entirely, I feel like I switched my mind on this because I heard this thing about, I think Einstein back in the day

Starting point is 00:28:50 where he's like, if you gave me 20 hours to solve a problem, I'd spend 19 hours defining it or something like that. And it's like kind of true, right? Like, it's so easy to go forward. in the right direction when you're sure that is the right direction, I guess. And like, I think almost everyone that I know, honestly, works really hard on like the first, like, it's not even the first direction. It's like the first, like, because you can, you can keep the same direction. It's just, like, I think the first angle or the first, like, starting point of that direction. Like, people just, people just immediately send it. And I think, like, you do get to a point where you are very, again, in tune with your intuition and you're able to really send it in the right direction at the

Starting point is 00:29:30 you know, in the correct place. But like, I just would implore people to, like, go to the club or something, like, do other things to, like, just keep thinking about your idea in different places. I think you'll realize how much of your past stuff has been, you know, just, I don't know how to describe it. Like, just not, you know, you don't want to think things through too much, right? but um yeah i one thing i've noticed is that like if i'm brewing an on an idea and i like when i step out of the house and i just like let it brew in the back burner yeah i'm so like kind of working but i'm not really like i'm on a walk or i'm riding my bike or doing something those moments i find

Starting point is 00:30:19 are just as key as the moments i'm in front of my computer exactly i think people need to get over this this anxiety of feeling that they're not working hard enough because they're not at their computer, right? Like, as you were just saying, I, uh, I feel like you as a founder are like always working, right? Like, I feel like I'm always working because I'm not, you don't really view your work as work, right? For the most part, but, um, you know, right now I am working in a sense and I'm still technically thinking through my ideas. Um, and I, I'm still open to change, you know, my perspectives and my ideas, no matter what I'm doing. And again, as most people are, the way you probably came up with your stupid startup idea in the first place is probably because you were like, you know,

Starting point is 00:31:00 in the shower, you were doing something else, right? And the idea came to you. Um, but as I was saying earlier, you know, a startup is a billion ideas anyways. You will go down a fat idea maze. Uh, I think it's important for you to, um, you know, just keep, keep thinking through some of these, these things in your mind. Yeah, I guess what I wanted to, to kind of bring to the table, I guess, kind of like a request for startups. I see that YC does this pretty often is, um, I'm obviously trying to build a chat bot that you form a relationship with. And I think this is like a category that's that's very popular right now, you know, regarding AI companionship. But I think like there's a broad use case of chatbots that aren't just like, you know, AI girlfriends, right? I think might be

Starting point is 00:31:42 what people are going in their minds immediately to. But, you know, there's a lot of people trying to build therapist bots, assistants, you know, mentor things, just fun little games, you know, toys, even kind of more what I'm doing, which is this kind of like embodied Tamagodians. friend type thing. Regardless, you know, to have a chat bot that you send more than a couple messages to, you want it to, you know, have a good memory system. And there's a couple ways people are trying to build memory systems these days. I think at the end of the day, it's really just rag, which is still honestly kind of a fake way to build a memory system, but it works for the time being and, you know, for the near future for sure. And what I really want to see is like memory

Starting point is 00:32:22 as a service, essentially. I want to see more people build tools that make it easier for you to extract memories from conversations and, you know, have a very good eval system for your own kind of conversations that happen with your bot to make sure it's, you know, generating those memories and, you know, a very quick low latency query system for these memories. You know, I think people would probably be worried about maybe offloading such a, such a crucial part of their application, but I honestly think it's enough of an issue that has just not been solve that people should really be focusing on this. I've I've taken a look now at nearly every way of doing this and every project out there. And I honestly think the only people doing a semi

Starting point is 00:33:08 good job of this is, I would say, dot from new computer. They've really nailed, like, the whole memory system because the trickiest part about this, I would say, is updating old memories and having everything be connected to each other, because very often I might, you know, well, for starters, you know, you're often changing your old memories. Like one day you might like a Ferrari, the next day you might like a Porsche, whatever, these things change. But more importantly, you want to be able to connect memories that are not semantically related to each other. Because I might say, I like, I don't know, I like cars, but somehow there is like a memory that's not really semantically related to cars or Ferrari or something like that. But it's still relevant to the conversation we're having right now.

Starting point is 00:33:54 like maybe, I don't know, like my girlfriend got in this car accident ages ago. And somehow that's like relevant. Basically, it's a really complicated problem. But, you know, I'm sure people can solve this. And I've taken a look online at some tools. And I think the only real open source projects that exist for this is like MGPT and MMMZero, which are okay, simple implementations. But, you know, I would, I think there's many people that would gladly pay for a service like this.

Starting point is 00:34:24 So I'd like to discuss that with you, Greg. Cool. So I think a lot of, you know, a bunch of people listening don't even know what the rag memory system is. So I'd love you to walk through. I'd actually really love you to walk through what that is, just like at a high level, because a lot of people aren't super technical. Yeah. And then how to think about like, is this a good idea or not?

Starting point is 00:34:51 Because there are competitors out there. Like, how do you know, how do you think about, like, oh my God, this is something someone should do and it would work. What do you mean competitors out there? Like, you mentioned two competitors, meaning, well, I don't know. They're building like, you know, they're building personalities. Like they're building these niche. I don't know, I'd say niche, but like, you know, they're building chatbots.

Starting point is 00:35:14 And like, honestly, if you're worried about competition, then that's your problem. You shouldn't be building a startup in the first place. I don't know what to tell you. You know, if you view your work as art, you have no competition. You have no fear or failure. that is entirely your problem. I'm not here to discuss your fears of competition. But RAG stands for, I believe, retrieval augmented generation, I think, which just means that,

Starting point is 00:35:36 you know, you give memory, like, just like facts or memories or whatever as part of, like, your system prompt or user prompt. And those memories obviously change the output, right? Like if I ask a question to, you know, you have a system prompt. that just says, like, I am a helpful assistant. And then I ask a question, you know, what is my favorite car? Obviously, that bot is not going to have any memory of that and is just going to say, well, I don't know, you never told me that.

Starting point is 00:36:06 But if you have a system where when I say, what is my favorite car, it's able to take basically what I said. And then I believe the way it kind of works right now is you would basically embed that. So you would kind of, I don't know how to explain embeddings necessarily. You kind of just create like a semantic, like relational map, I guess, of like that sentence that I said. So, you know, you would basically just take like car, right, and find something that, you know, is the closest match to car, right? But you would, you would take that. Well, I guess before you would do that, you would say to the assistant, you know, my favorite car is a Porsche, okay?

Starting point is 00:36:49 A Porsche 9-11. And it would just kind of, you know, the models have a car. concept of automobiles and cars, right? And like a Porsche 9-11 is like kind of close to that concept of a car. So if you imagine just like a 3D space, right, those are, they're kind of just close to each other. Okay. And then so later on, um, if I ask like, what is my favorite car? It'll, you know, see the word car and like look for something kind of similar. And, you know, it's not going to be a perfect match, right? You know, Porsche 9-11 is not the word car, but it's close enough. Maybe it's like a 0.8.9 match, right?

Starting point is 00:37:24 Something like that, like a 90% match. So it's close enough. So you would just choose like some of these closest memories. Right now, let's say we only have that one memory. We would take that memory and we might put that in, let's say the, I don't know, let's say the user message that gets sent to this, this bot where, you know, the system promise I'm a helpful assistant. Then your user message is what is my favorite car?

Starting point is 00:37:46 Then maybe you have underneath that, you know, you augment this message and you say like, you know, in brackets, like the user's favorite, the user's favorite car is a Porsche 9-11, or you maybe don't even need to, you know, augment the way that's written. You can just say my user chat message that I sent ages ago, like, my favorite car is a Porsche 9-11, right? And so when the system prompt sees, you know, your question and then, you know, the little context that you have there, it'll be able to say like, oh, you know, your favorite car is a portion 9-11, right? So that's like, kind of how it works. works, you would kind of store all of these memories in probably some kind of like vector database,

Starting point is 00:38:26 right? You don't really need to use a standalone tool. A lot of people are probably using Superbase as their database and you can just kind of, you know, use like PG vector in there. They make it all pretty simple. But, you know, to set up a system like this that can extract those memories and put those memories back into your chats alone is already, you know, pretty technical. but then again, building a true memory system, like I was mentioning earlier, your memory is a lot more adaptive, right? Like, you know, you need to update existing memories. What if now my favorite car is a, you know, I don't know, a Pagani, something, whatever, right? Now that needs to be updated.

Starting point is 00:39:03 And if you think about the way your brain's working too, right? Like, you have, like, the importance of these memories matters a lot. And that will degrade over time as well. You know, recency matters a lot for some of these memories. because if it's trying to maybe, maybe you're trying to build a system that updates existing memories. And maybe I said my favorite car was a Porsche 9-11, but that was five years ago or five months ago, whatever. And only yesterday that I said that my favorite car was, you know, I guess I don't know. You know, fuck cars. I like motorcycles more.

Starting point is 00:39:38 I'd say like a Dakota 916, you know, is now my favorite, my favorite transportation thing, right? Like, you need to update that memory, which means that you really need to have, like, a back-end system for doing this. And I think, like, the real way that you would build this is with, like, a graph network, right? Which gets pretty complicated. But that way you can fix this issue I mentioned earlier of connecting non-semantically related things with each other. Like, in a sense, you know, you probably would want to include that I like a Dukadi 916, even if I ask, like, what is my favorite car, right? but, you know, Ducati is a motorcycle. It's not a car. They're not inherently connected. You know, it still might be, let's say how I mentioned earlier. You know, if I mentioned Portia 9-11,

Starting point is 00:40:23 that is like, you know, a 0.8.9 similarity, right? And the Ducati is still generally closer to a car than if I mentioned, you know, a horse, right? This is my favorite color horse. That's obviously going to be maybe like, you know, 0.01, let's say. It's just so far away. But a Jakati 916 is still may be going to be like, you know, point four, point three, let's say. But yeah, does that work for you, Greg? Yeah, I think, okay, so I'll be honest. I don't feel personally like I would be able to execute this particular idea and win. That's because I'm just not super technical, but I can, I'm technical, I'm technical enough

Starting point is 00:41:03 to hear what you're saying and be like, that's a real problem. someone is going to do really well to fix it. And the time is now, you know? Oh, for sure. I mean, I don't know how much longer, you know, rag is necessarily going to be an issue. But I know for sure that memory is going to be something that all of these bots need.

Starting point is 00:41:23 And, you know, I think it is, again, pretty broad because I know a billion people, you know, there's a lot of people trying to build chatbots and a lot of these people need memory systems. And I have looked at all of these people's startups and I know that everyone is struggling with memory as well. No one has truly cracked. Even D.DOT, I mentioned earlier from a new computer.

Starting point is 00:41:41 They get pretty close, but, you know, nothing's perfect. I think memory is essential for bringing users back into, you know, having high retention with any of these bots. So it is, you know, a top priority. And I think any founder that is stuck with, you know, their memory system would go to great length to make sure it works really well, you know, myself included. This is a request for, you know, I need help myself. Totally. It gets me thinking that if you want to build a startup, maybe one, and you're looking for startup ideas that are going to work, maybe a good way is to just ask a particular type of founder, i.e. AI hardware founders or e-commerce founders and be like, what are the five biggest problems that you would pay money for to solve? What do you think about that?

Starting point is 00:42:27 Yeah, I mean, honestly, this is the problem I would pay the most money to solve. It is my real problem. Like, you know, I've got a team and we are, working on it and I've been thinking about this for a while. I think the reason why I'm bringing it up as maybe a very strong priority is I have a lot more context and memories to deal with than, you know, just a web-based chat bot given that I have an always-on microphone that is listening to everything you say. And that's a lot of, that's just a lot of data to parse through, especially because think of how often you change your mind, right, in real life. And if you're listening to those and trying to form memories off of that. Well, man, I mean, you need a really robust memory system.

Starting point is 00:43:07 I think, you know, it's great that all these great ideas are hampered by, you know, these are kind of just like simple technical problems that I think will be solved. But yeah. And I guess it is a technical problem, but I think it's more, I mean, I don't know what to tell you. I mean, skill issue. Like, every real idea is going to be technical. Every real idea is going to be technical. What do you mean by that?

Starting point is 00:43:34 I mean, I can't think of any startup ideas that would be inherently. I mean, like, okay, like, no matter what you're working on, even if it's like a shovel, you know, a really good shovel, somebody is going to need to engineer it. And like, sure, your role, if you're listening to this, might be, okay, you really know how to pitch this as a problem and you find some engineers that you hire and, you know, your job is, you know, you raise funding and everything like that. You know, I'm not necessarily saying that you.

Starting point is 00:43:59 need to be the technical person working on it. But I mean, yeah, like I'm looking at a billion objects around me, and I don't think anything of this is not inherently technical to have built in the first place. You don't think that business creation has like a spectrum of how technical things are. So, for example, you know, Morning Brew, for example, is, I think they make like $100 million a year or something like that. and but it's a newsletter business. They're not, they're just built on top of,

Starting point is 00:44:32 you know, existing infrastructure versus, you know, building this memory idea. To me, technical is not like technology, like high technology necessarily, you know,

Starting point is 00:44:43 like programming, okay, like I'm looking around me at, there's a whiteboard over there, there's an air purifier, there's a rug over here, a water bottle, an SD card,

Starting point is 00:44:51 a rock that's shaped like a pyramid. I don't know, all these things, are inherently technical in many aspects. And like, even if you are business oriented, right, like, you still, there's still technical aspects to setting up a business and, you know, pitching in a sense. Like, these are, everything is, I mean, technical just, I don't even know what the definition of that is, just like, like, details. I don't know how to describe it. Just, no, I love, I love hearing, because, you know, it's different than how I think, you know? So,

Starting point is 00:45:26 And I, you know, it's part of the pod. Like, I want to bring on diversity of people, right? So if, I just want to make sure I was understanding you correctly. Basically, what you're saying is all businesses are technical. Therefore, in order to be, like, you need to be. Let me, let me just like, like, all businesses are technical because business is technical. Is it not? I think that some businesses are more technical than other businesses.

Starting point is 00:45:52 So if you're going to go in, like, if I were to go to create Varda, which. which is like building like, sure, warehouses in space or whatever. I think that's way harder than building a newsletter, for example. I do agree with you. Sure. Sure.

Starting point is 00:46:09 I think, yeah, I don't think like startup success also is correlated with like how technical, you know, your thing is, right? Like, let's say,

Starting point is 00:46:16 Warning Brew, I guess you were mentioning is not, you know, inherently technical and maybe the same way that a lot of people perceive that definition. But, um, they,

Starting point is 00:46:26 you know, might still succeed a lot more than someone that's building something so high-tech, like someone that's trying to build fusion reactors, let's say, or something really deep tech and weird and crazy, or some mad scientist trying to come up with some weird idea. Yeah, you know, tech-technical does not necessarily correlate with success. And maybe I think that's the path some people might mistakenly go down. You know, they think it needs to be more complicated than it really needs to be. But isn't the whole point of engineering, too, to simplify,

Starting point is 00:46:55 complexity anyways. Totally. I think that's a beautiful place to end. Sure. Yeah. Any, anything, any last words for people and where could people get to follow your thoughts? I don't know. Just don't be boring and don't work on something stupid. And again, you're only going to work on so many things in your life. You know, you have a short bubble of consciousness while we're alive right now. Who knows? Maybe we'll extend that during our lifetimes. But, you know, in the short bubble of consciousness. Just don't waste it's something you don't feel confidently proud about. And if you're, if you are, then I don't know. I think that's just your problem and no one else is going to tell you anything otherwise. In terms of where you can find me, definitely friend.com for now or if you have

Starting point is 00:47:45 some pop-up thingy, I mean, sure, you can go to my Twitter. I would highly recommend if you want to see me post videos about motorcycles for the most part. But yeah, thank you. Thank you, Greg. This is That would be great. I think this is, we'll go down is probably, we've never had an episode like this. Let's put it that way. We've never had an episode like this. So the common section will be buzzing. And I think some people will be excited by what you say.

Starting point is 00:48:15 And then some people will be like, I don't want that. Yeah. Yeah. I guess I would implore people that are familiar with who I am and what I'm working on to see that this is why I am completely unaffected by what you have to say about what I have to say. For the most part, you know, I've been in the media now for like four or five years. And with a specifically friend, I will say that it has significantly uptick the amount of people that just do not like who I am and do not like what I'm working on. I get it's a very controversial industry.

Starting point is 00:48:44 But 100% the way I view a lot of people's feedback is it allows me to experience my own artwork from a different person's perspective. And I love that. So, you know, I implore people to view it from that angle. But thank you, Greg. Thank you, Avi. Appreciate it. See ya. See ya.

The Startup Ideas Podcast - "Y Combinator Is Overrated" - Inside the brain of this 20-Year-Old Who Built Multiple Viral Products

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.